Farzana Rahman
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Farzana Rahman has a total of 247 co-authors. Here are the top co-authors by number of shared publications:
- Bart Cuypers
ADReM Data Lab, Department of Computer Science University of Antwerp 2020 Antwerp Belgium
6 shared publications - Dan DeBlasio
Ray and Stephanie Lane Computational Biology Department, Carnegie Mellon University, Pittsburgh, PA 15213, United States
6 shared publications - Jean-Christophe Nebel
School of Computing and Mathematics, Faculty of Engineering, Computing, and Environment, Kingston University, London, KT1 2EE, United Kingdom
6 shared publications - Mehedi Hassan
School of Computing and Mathematics, University of South Wales, Pontypridd, United Kingdom
6 shared publications - Rodrigo Gonzalo Parra
Evolutionary Systems Biophysics Group, Life Sciences Department, Barcelona Supercomputing Center, Barcelona, Spain
6 shared publications - Abdulsamie Hanano
Department of Molecular Biology and Biotechnology, Atomic Energy Commission of Syria (AECS), Damascus, Syria
5 shared publications
Latest Publications
- Open Access
QUARTZ: Cross-Dataset Artificial Intelligence-Enabled Retinal Vasculometry
Examination of retinal images provides a non-invasive means to assess the body’s microvascular system, with vessel morphology linked to ocular and systemic diseases. QUARTZ (QUantitative Analysis of Retinal vessel Topology and siZe) is an automated, artificial intelligence-enabled retinal vasculometry system designed for processing large-scale retinal image datasets, including over 4 million images from within the National Health Service, to facilitate epidemiological studies. The system is uniquely capable of extracting useful information from lower quality images, including those which are partially illuminated, enabling analysis of a greater proportion of images. To enhance cross-dataset generalisation of the system from an earlier dataset-tailored approach, this work leverages 26 annotated retinal image datasets spanning diverse demographics, diseases, camera types, and imaging protocols. Evaluation on the external EPIC-Norfolk dataset demonstrates robust performance with F1 scores of 0.8305, 0.8477, 0.8741, 0.9532, and 0.8677 for arteriole, venule, vessel, optic disc, and optic cup segmentation, respectively, and 0.9197 for image quality classification. The cross-dataset generalisation approach demonstrated increased performance for arteriole/venule segmentation and optic disc segmentation. However, improvements in performance were not observed for optic cup segmentation, or for image quality classification. An online tool is provided to support wider adoption and reproducible biomarker research across diverse populations and imaging settings. In addition, the new public CHASE-AV database is introduced for arteriole and venule segmentation.
Open AccessQUARTZ: Cross-Dataset Artificial Intelligence-Enabled Retinal Vasculometry
Examination of retinal images provides a non-invasive means to assess the body’s microvascular system, with vessel morphology linked to ocular and systemic diseases. QUARTZ (QUantitative Analysis of Retinal vessel Topology and siZe) is an automated, artificial intelligence-enabled retinal vasculometry system designed for processing large-scale retinal image datasets, including over 4 million images from within the National Health Service, to facilitate epidemiological studies. The system is uniquely capable of extracting useful information from lower quality images, including those which are partially illuminated, enabling analysis of a greater proportion of images. To enhance cross-dataset generalisation of the system from an earlier dataset-tailored approach, this work leverages 26 annotated retinal image datasets spanning diverse demographics, diseases, camera types, and imaging protocols. Evaluation on the external EPIC-Norfolk dataset demonstrates robust performance with F1 scores of 0.8305, 0.8477, 0.8741, 0.9532, and 0.8677 for arteriole, venule, vessel, optic disc, and optic cup segmentation, respectively, and 0.9197 for image quality classification. The cross-dataset generalisation approach demonstrated increased performance for arteriole/venule segmentation and optic disc segmentation. However, improvements in performance were not observed for optic cup segmentation, or for image quality classification. An online tool is provided to support wider adoption and reproducible biomarker research across diverse populations and imaging settings. In addition, the new public CHASE-AV database is introduced for arteriole and venule segmentation.
- 20 July 2026
- Published by Springer Nature
- p. 74-88
Adaptive and Uncertainty-Aware Intrusion Detection with Hybrid Meta-learning Under Proxy-Adversarial Exposure
- 20 July 2026
- Published by Springer Nature
- p. 74-88
Adaptive and Uncertainty-Aware Intrusion Detection with Hybrid Meta-learning Under Proxy-Adversarial Exposure
- Open Access
- 1 July 2026
- Published by Oxford University Press (OUP) in Bioinformatics
- Vol. 42 (Supplement)
Quantifying domain-specific relevance of computational biology Wikipedia articles using TF-IDF and cosine similarity.
Wikipedia is one of the world's most visited websites and serves as the principal open educational resource for computational biology. However, identifying which articles are most relevant to distinct sub-disciplines of computational biology remains largely subjective. This study collected short descriptions for 22 Communities of Special Interest (COSI) groups maintained by the International Society for Computational Biology and downloaded 1536 computational biology articles from English Wikipedia. Following standard text preprocessing, COSI descriptions and Wikipedia articles were embedded in a common TF-IDF vector space. Semantic relatedness was quantified using cosine similarity, yielding a real-valued relevance matrix that maps each COSI to the most pertinent computational biology articles. The resulting scores, typically low in absolute value, captured nuanced differences: general-interest pages such as 'Computational biology' and 'Bioinformatics' ranked highest, whereas niche pages showed high relevance only for specific COSIs. Unsupervised analysis using principal component analysis, k-nearest neighbours, and Leiden community detection revealed clusters of articles corresponding to the particular COSIs and highlighted inter-COSI relationships. This automated pipeline reduces bias compared with manual tagging and enables more precise curation of domain-specific educational resources. The relevance matrix developed in this study is available in the Zenodo repository (doi: 10.5281/zenodo.18311878).
- 1 July 2026
- Published by Oxford University Press (OUP) in Bioinformatics
- Vol. 42 (Supplement)
Open AccessQuantifying domain-specific relevance of computational biology Wikipedia articles using TF-IDF and cosine similarity.
Wikipedia is one of the world's most visited websites and serves as the principal open educational resource for computational biology. However, identifying which articles are most relevant to distinct sub-disciplines of computational biology remains largely subjective. This study collected short descriptions for 22 Communities of Special Interest (COSI) groups maintained by the International Society for Computational Biology and downloaded 1536 computational biology articles from English Wikipedia. Following standard text preprocessing, COSI descriptions and Wikipedia articles were embedded in a common TF-IDF vector space. Semantic relatedness was quantified using cosine similarity, yielding a real-valued relevance matrix that maps each COSI to the most pertinent computational biology articles. The resulting scores, typically low in absolute value, captured nuanced differences: general-interest pages such as 'Computational biology' and 'Bioinformatics' ranked highest, whereas niche pages showed high relevance only for specific COSIs. Unsupervised analysis using principal component analysis, k-nearest neighbours, and Leiden community detection revealed clusters of articles corresponding to the particular COSIs and highlighted inter-COSI relationships. This automated pipeline reduces bias compared with manual tagging and enables more precise curation of domain-specific educational resources. The relevance matrix developed in this study is available in the Zenodo repository (doi: 10.5281/zenodo.18311878).
- Open Access
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-5
Bridging Factions: A Meet-up Connecting Diverse Perspectives in Human–Computer Interaction
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-5
Open AccessBridging Factions: A Meet-up Connecting Diverse Perspectives in Human–Computer Interaction
- Open Access
Reaping the Fruits of LLM Pruning: Towards Small Language Models for Efficient Non-Coding Variant Effect Prediction
Background: Interpreting variant effects is essential for precision medicine. Large Transformer-based genomic language models (DNABERT 2, Nucleotide Transformer) capture patterns in coding DNA but scale poorly for non coding variant prediction because attention complexity grows quadratically with sequence length. Evidence from natural language processing shows that pruning less informative layers can reduce model size and computational load without sacrificing accuracy. Methods: We systematically ablated each Transformer layer in DNABERT 2 and the Nucleotide Transformer to assess its contribution to variant prediction. By observing changes in performance, we built layer importance profiles and created pruned models by removing redundant layers. Pruned and full models were fine tuned with identical hyperparameters using the Enformer eQTL causal variant dataset, a curated benchmark for non coding variant effect prediction. Results: Layer ablation revealed that the importance of individual layers varies widely across models; some layers can be removed with little loss in performance while others are critical. After fine tuning, pruned models achieved accuracy and area under the ROC curve comparable to full models. Additionally, pruned versions required substantially less training time and memory, reducing resource usage by a significant margin. Conclusions: Layer wise pruning provides a principled strategy for developing compact genomic LLMs. By identifying and removing less critical layers, we produced leaner models that preserve predictive power while lowering computational demands. These efficient models demonstrate how insights from general LLM research can advance genomic variant interpretation and make large scale non coding analysis more accessible in research and clinical settings. This approach complements ongoing efforts to optimise Transformer architectures for genomic data.
Open AccessReaping the Fruits of LLM Pruning: Towards Small Language Models for Efficient Non-Coding Variant Effect Prediction
Background: Interpreting variant effects is essential for precision medicine. Large Transformer-based genomic language models (DNABERT 2, Nucleotide Transformer) capture patterns in coding DNA but scale poorly for non coding variant prediction because attention complexity grows quadratically with sequence length. Evidence from natural language processing shows that pruning less informative layers can reduce model size and computational load without sacrificing accuracy. Methods: We systematically ablated each Transformer layer in DNABERT 2 and the Nucleotide Transformer to assess its contribution to variant prediction. By observing changes in performance, we built layer importance profiles and created pruned models by removing redundant layers. Pruned and full models were fine tuned with identical hyperparameters using the Enformer eQTL causal variant dataset, a curated benchmark for non coding variant effect prediction. Results: Layer ablation revealed that the importance of individual layers varies widely across models; some layers can be removed with little loss in performance while others are critical. After fine tuning, pruned models achieved accuracy and area under the ROC curve comparable to full models. Additionally, pruned versions required substantially less training time and memory, reducing resource usage by a significant margin. Conclusions: Layer wise pruning provides a principled strategy for developing compact genomic LLMs. By identifying and removing less critical layers, we produced leaner models that preserve predictive power while lowering computational demands. These efficient models demonstrate how insights from general LLM research can advance genomic variant interpretation and make large scale non coding analysis more accessible in research and clinical settings. This approach complements ongoing efforts to optimise Transformer architectures for genomic data.
- 1 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Biology and Bioinformatics
- Vol. 22 (6) , 2575-2585
A Deep Learning Framework for Protein-to-Metal Binding Prediction Using Protein Language Models
This study presents an end-to-end deep learning framework for protein-to-metal-ion binding prediction, a critical task in understanding protein function, structural stability, and metal transport mechanisms. A binding site is a residue location in a protein sequence where a metal binds to a protein. Manual curation of metal binding sites is a tedious process involving mining through research articles, making it expensive, laborious, and time-consuming. Therefore, developing a computational pipeline is essential to predict metal ion binding of unannotated proteins. A significant shortcoming of existing computational methods is the failure to capture the long-term dependency of the residues, the absence of positional information, and a pre-determined set of residues and metal ions. In this paper, we propose a metal-ion binding prediction pipeline using a large language model, emphasizing 1) the comparative performance of five state-of-the-art protein language models (pLMs), 2) the impact of positional encoding of binding sites, and 3) the comparison with classical machine learning techniques. A 10-fold cross-validation evaluation yielded a Matthews Correlation Coefficient (MCC) of 0.89, along with precision, recall, and F1 scores exceeding 95% for the six most extensively studied metal ions reported in the literature.
- 1 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Biology and Bioinformatics
- Vol. 22 (6) , 2575-2585
A Deep Learning Framework for Protein-to-Metal Binding Prediction Using Protein Language Models
This study presents an end-to-end deep learning framework for protein-to-metal-ion binding prediction, a critical task in understanding protein function, structural stability, and metal transport mechanisms. A binding site is a residue location in a protein sequence where a metal binds to a protein. Manual curation of metal binding sites is a tedious process involving mining through research articles, making it expensive, laborious, and time-consuming. Therefore, developing a computational pipeline is essential to predict metal ion binding of unannotated proteins. A significant shortcoming of existing computational methods is the failure to capture the long-term dependency of the residues, the absence of positional information, and a pre-determined set of residues and metal ions. In this paper, we propose a metal-ion binding prediction pipeline using a large language model, emphasizing 1) the comparative performance of five state-of-the-art protein language models (pLMs), 2) the impact of positional encoding of binding sites, and 3) the comparison with classical machine learning techniques. A 10-fold cross-validation evaluation yielded a Matthews Correlation Coefficient (MCC) of 0.89, along with precision, recall, and F1 scores exceeding 95% for the six most extensively studied metal ions reported in the literature.
- Open Access
- 15 October 2025
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 21 (10) , e1013554
Impact of the COVID-19 pandemic on computational biology early career researchers: A global retrospective study
The COVID-19 pandemic led to devastating physical, psychological, and financial impacts on millions of people across the world. Amidst a rapidly evolving research landscape, the global scientific community was forced to swiftly adapt to novel working methods, including remote collaboration tools, virtual conferences, and online research platforms. Surveys of life sciences researchers have indicated that computational biologists experienced less disruption and a smoother transition to remote working than experimental biologists, due to their reduced reliance on laboratory equipment. Despite this adaptability, the sudden shift to remote work, compounded by stress and social isolation, has posed significant mental health challenges for these workers. However, remote work has also facilitated opportunities for more flexible work arrangements and increased collaboration across geographical boundaries. To investigate these impacts, we conducted surveys of computational biologists during the Intelligent Systems for Molecular Biology (ISMB) conferences in 2020 and 2021, which were held virtually due to the COVID-19 lockdowns. This study implements a thorough statistical analysis of the survey results to offer insights into the repercussions of the lockdowns on researchers and their work. Key areas of investigation include the effects of institutional support (or lack thereof), the difference in productivity compared to pre-lockdown levels, and the significance of gender in determining these impacts. Notably, a lack of institutional support with regard to mental health and finances was shown to have a significant negative effect on early-career researchers. Although limited by a small sample size, our study sets the stage for a more robust exploration of these trends in future research. Importantly, by illuminating the challenges and opportunities arising from the COVID-19 pandemic and lockdowns, our study offers hope for potential solutions supporting the well-being of early-career researchers in unprecedented circumstances.
- 15 October 2025
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 21 (10) , e1013554
Open AccessImpact of the COVID-19 pandemic on computational biology early career researchers: A global retrospective study
The COVID-19 pandemic led to devastating physical, psychological, and financial impacts on millions of people across the world. Amidst a rapidly evolving research landscape, the global scientific community was forced to swiftly adapt to novel working methods, including remote collaboration tools, virtual conferences, and online research platforms. Surveys of life sciences researchers have indicated that computational biologists experienced less disruption and a smoother transition to remote working than experimental biologists, due to their reduced reliance on laboratory equipment. Despite this adaptability, the sudden shift to remote work, compounded by stress and social isolation, has posed significant mental health challenges for these workers. However, remote work has also facilitated opportunities for more flexible work arrangements and increased collaboration across geographical boundaries. To investigate these impacts, we conducted surveys of computational biologists during the Intelligent Systems for Molecular Biology (ISMB) conferences in 2020 and 2021, which were held virtually due to the COVID-19 lockdowns. This study implements a thorough statistical analysis of the survey results to offer insights into the repercussions of the lockdowns on researchers and their work. Key areas of investigation include the effects of institutional support (or lack thereof), the difference in productivity compared to pre-lockdown levels, and the significance of gender in determining these impacts. Notably, a lack of institutional support with regard to mental health and finances was shown to have a significant negative effect on early-career researchers. Although limited by a small sample size, our study sets the stage for a more robust exploration of these trends in future research. Importantly, by illuminating the challenges and opportunities arising from the COVID-19 pandemic and lockdowns, our study offers hope for potential solutions supporting the well-being of early-career researchers in unprecedented circumstances.
- Open Access
- 1 July 2025
- Published by Oxford University Press (OUP) in Bioinformatics
- Vol. 41 (Supplement) , i12-i20
An educator framework for organizing Wikipedia editathons for computational biology
Motivation: Wikipedia is a vital open educational resource in computational biology; however, a significant knowledge gap exists between English and non-English Wikipedias. Reducing this knowledge gap via intensive editing events, or “editathons,” would be beneficial in reducing language barriers that disadvantage learners whose native language is not English. Results: We present a framework to guide educators in organizing editathons for learners to improve and create relevant Wikipedia articles. As a case study, we present the results of an editathon held at the 2024 ISCB Latin America conference, in which ten new articles were created for the Spanish-language edition of Wikipedia. We also present a web tool, “compbio-on-wiki,” which identifies relevant English Wikipedia articles missing in other languages. We demonstrate the value of editathons to expand the accessibility and visibility of computational biology content in multiple languages. Availability and implementation: Source code for the compbio-on-wiki Toolforge site is available at: https://github.com/lubianat/compbio-on-wiki
- 1 July 2025
- Published by Oxford University Press (OUP) in Bioinformatics
- Vol. 41 (Supplement) , i12-i20
Open AccessAn educator framework for organizing Wikipedia editathons for computational biology
Motivation: Wikipedia is a vital open educational resource in computational biology; however, a significant knowledge gap exists between English and non-English Wikipedias. Reducing this knowledge gap via intensive editing events, or “editathons,” would be beneficial in reducing language barriers that disadvantage learners whose native language is not English. Results: We present a framework to guide educators in organizing editathons for learners to improve and create relevant Wikipedia articles. As a case study, we present the results of an editathon held at the 2024 ISCB Latin America conference, in which ten new articles were created for the Spanish-language edition of Wikipedia. We also present a web tool, “compbio-on-wiki,” which identifies relevant English Wikipedia articles missing in other languages. We demonstrate the value of editathons to expand the accessibility and visibility of computational biology content in multiple languages. Availability and implementation: Source code for the compbio-on-wiki Toolforge site is available at: https://github.com/lubianat/compbio-on-wiki
- Open Access
Air Pollution Monitoring Using Cost-Effective Devices Enhanced by Machine Learning
Given the significant impact of air pollution on global health, the continuous and precise monitoring of air quality in all populated environments is crucial. Unfortunately, even in the most developed economies, current air quality monitoring networks are largely inadequate. The high cost of monitoring stations has been identified as a key barrier to widespread coverage, making cost-effective air quality monitoring devices a potential game changer. However, the accuracy of the measurements obtained from low-cost sensors is affected by many factors, including gas cross-sensitivity, environmental conditions, and production inconsistencies. Fortunately, machine learning models can capture complex interdependent relationships in sensor responses and thus can enhance their readings and sensor accuracy. After gathering measurements from cost-effective air pollution monitoring devices placed alongside a reference station, the data were used to train such models. Assessments of their performance showed that models tailored to individual sensor units greatly improved measurement accuracy, boosting their correlation with reference-grade instruments by up to 10%. Nonetheless, this research also revealed that inconsistencies in the performance of similar sensor units can prevent the creation of a unified correction model for a given sensor type.
Open AccessAir Pollution Monitoring Using Cost-Effective Devices Enhanced by Machine Learning
Given the significant impact of air pollution on global health, the continuous and precise monitoring of air quality in all populated environments is crucial. Unfortunately, even in the most developed economies, current air quality monitoring networks are largely inadequate. The high cost of monitoring stations has been identified as a key barrier to widespread coverage, making cost-effective air quality monitoring devices a potential game changer. However, the accuracy of the measurements obtained from low-cost sensors is affected by many factors, including gas cross-sensitivity, environmental conditions, and production inconsistencies. Fortunately, machine learning models can capture complex interdependent relationships in sensor responses and thus can enhance their readings and sensor accuracy. After gathering measurements from cost-effective air pollution monitoring devices placed alongside a reference station, the data were used to train such models. Assessments of their performance showed that models tailored to individual sensor units greatly improved measurement accuracy, boosting their correlation with reference-grade instruments by up to 10%. Nonetheless, this research also revealed that inconsistencies in the performance of similar sensor units can prevent the creation of a unified correction model for a given sensor type.
- Open Access
- 5 February 2025
- Published by MDPI AG in Preprints.org
Air Pollution Monitoring using Cost-Effective Devices
Given the significant impact of air pollution on global health, continuous and precise monitoring of air quality in all populated environments is crucial. Unfortunately, even in the most developed economies, current air quality monitoring networks are largely inadequate. The high cost of monitoring stations has been identified as a key barrier to widespread coverage, making cost-effective air quality monitoring devices a potential game changer. However, the accuracy of measurements obtained from low-cost sensors is affected by many factors, including gas cross-sensitivity, environmental conditions, and production inconsistencies. Fortunately, machine learning models can capture complex interdependent relationships in sensor responses, and thus can enhance their readings and sensor accuracy. After gathering measurements from cost-effective air pollution monitoring devices placed alongside a reference station, the data were used to train such models. Assessment of their performance showed that models tailored to individual sensor units greatly improved measurement accuracy, boosting their correlation with reference-grade instruments by up to 10\%. Nonetheless, the research also revealed that inconsistencies in the performance of similar sensor units can prevent the creation of a unified correction model for a given sensor type.
- 5 February 2025
- Published by MDPI AG in Preprints.org
Open AccessAir Pollution Monitoring using Cost-Effective Devices
Given the significant impact of air pollution on global health, continuous and precise monitoring of air quality in all populated environments is crucial. Unfortunately, even in the most developed economies, current air quality monitoring networks are largely inadequate. The high cost of monitoring stations has been identified as a key barrier to widespread coverage, making cost-effective air quality monitoring devices a potential game changer. However, the accuracy of measurements obtained from low-cost sensors is affected by many factors, including gas cross-sensitivity, environmental conditions, and production inconsistencies. Fortunately, machine learning models can capture complex interdependent relationships in sensor responses, and thus can enhance their readings and sensor accuracy. After gathering measurements from cost-effective air pollution monitoring devices placed alongside a reference station, the data were used to train such models. Assessment of their performance showed that models tailored to individual sensor units greatly improved measurement accuracy, boosting their correlation with reference-grade instruments by up to 10\%. Nonetheless, the research also revealed that inconsistencies in the performance of similar sensor units can prevent the creation of a unified correction model for a given sensor type.